Cloud based machine learning

Abstract

Disclosed are various embodiments for implementing computational tasks in a cloud environment in one or more operating system level virtualized containers. A parameter file can specify different parameters including hardware parameters, library parameters, user code parameters, and job parameters (e.g., sets of hyperparameters). The parameter file can be converted via a mapping and implemented in a cloud-based container platform.

What is claimed is: 
     
       1. A method comprising:
 receiving, using one more processors of a machine, a request to initiate one or more containers to run on a container platform, the request identifying a machine learning scheme parameter file that specifies model data for a machine learning model, platform parameters, and job-specific configurations for the one or more containers; 
 converting the machine learning scheme parameter file into a configuration file using a mapping template, the configuration file comprising the platform parameters and job-specific configurations native to the container platform; 
 initiating the one or more containers on the container platform with the configuration file by translating the model data into platform parameters native to the container platform; and 
 storing output data generated by the one or more containers. 
 
     
     
       2. The method of  claim 1 , wherein converting further comprises:
 using the mapping template to convert the machine learning scheme parameter file into the configuration file by mapping the platform parameters of the machine learning scheme parameter file into commands native to the container platform. 
 
     
     
       3. The method of  claim 1 ,
 wherein the machine learning scheme parameter file includes a detailed mapping of computational resources to specific machine learning tasks within the one or more containers, and wherein the initiating includes allocating the computational resources to the containers in accordance with the detailed mapping. 
 
     
     
       4. The method of  claim 1 , wherein the machine learning scheme parameter file specifies a container image for the machine learning model, wherein the container platform is configured to manage the one or more containers. 
     
     
       5. The method of  claim 4 , wherein the initiation is dynamically configured based on the platform parameters and job-specific configurations derived from the machine learning scheme parameter file. 
     
     
       6. The method of  claim 4 , wherein the model data comprises a network address of code that is executable in the one or more containers using the container image, the network address of the code located on a network server. 
     
     
       7. The method of  claim 4 , wherein the model data comprises machine learning model configuration data that specifies a set of configuration parameters of the machine learning model, wherein the machine learning model configuration data specifies a plurality of container jobs. 
     
     
       8. The method of  claim 7 , wherein each container job comprises a differently configured set of configuration parameters of the machine learning model. 
     
     
       9. The method of  claim 7 , wherein the initiating the one or more containers comprises initiating a plurality of containers using the same machine learning container image. 
     
     
       10. The method of  claim 4 , wherein the model data comprises hardware resource parameter data specifying one or more processor units and memory units, wherein the hardware resource parameter data specifies hardware resources for each container job. 
     
     
       11. A system comprising:
 one or more processors of the system; and 
 a memory storing instructions that, when executed by the one or more processors, cause the system to perform operations comprising: 
 receiving, using the one more processors, a request to initiate one or more containers to run on a container platform, the request identifying a machine learning scheme parameter file that specifies model data for a machine learning model, platform parameters, and job-specific configurations for the one or more containers; 
 converting the machine learning scheme parameter file into a configuration file using a mapping template, the configuration file comprising the platform parameters and job-specific configurations native to the container platform; 
 initiating the one or more containers on the container platform with the configuration file by translating the model data into platform parameters native to the container platform; and 
 storing output data generated by the one or more containers. 
 
     
     
       12. The system of  claim 11 , wherein converting further comprises:
 using the mapping template to convert the machine learning scheme parameter file into the configuration file. 
 
     
     
       13. The system of  claim 11 , wherein converting further comprises:
 mapping the platform parameters of the machine learning scheme parameter file into commands native to the container platform. 
 
     
     
       14. The system of  claim 11 , wherein the machine learning scheme parameter file specifies a container image for the machine learning model, wherein the container platform is configured to manage the one or more containers. 
     
     
       15. The system of  claim 14 , wherein the initiation is dynamically configured based on the platform parameters and job-specific configurations derived from the machine learning scheme parameter file. 
     
     
       16. The system of  claim 14 , wherein the model data comprises a network address of code that is executable in the one or more containers using the container image, the network address of the code located on a network server. 
     
     
       17. The system of  claim 14 , wherein the model data comprises machine learning model configuration data that specifies a set of configuration parameters of the machine learning model, wherein the machine learning model configuration data specifies a plurality of container jobs. 
     
     
       18. The system of  claim 17 , wherein each container job comprises a differently configured set of configuration parameters of the machine learning model. 
     
     
       19. The system of  claim 17 , wherein the initiating the one or more containers comprises initiating a plurality of containers using the same machine learning container image. 
     
     
       20. A non-transitory machine-readable storage device embodying instructions that, when executed by a device, cause the device to perform operations comprising:
 receiving, using one more processors of a machine, a request to initiate one or more containers to run on a container platform, the request identifying a machine learning scheme parameter file that specifies model data for a machine learning model, platform parameters, and job-specific configurations for the one or more containers; 
 converting the machine learning scheme parameter file into a configuration file using a mapping template, the configuration file comprising the platform parameters and job-specific configurations native to the container platform; 
 initiating the one or more containers on the container platform with the configuration file by translating the model data into platform parameters native to the container platform; and 
 storing output data generated by the one or more containers, wherein the output data includes metrics indicative of the execution of the containers according to the job-specific configurations.